The physicist envisioned robots replacing human workers
The proverb “success has many fathers” is rarely clearer than in the many stories about the rise of Silicon Valley and the diverse web of people, institutions, resources, and dynamics that transformed the San Francisco Peninsula into an astonishingly intense locus of technological activity. Several threads of this web are now legend: Dave Packard and Bill Hewlett’s garage workshop, the tireless networking of Stanford engineering dean Fred Terman, the audacious young scientists and engineers who founded Fairchild Semiconductor Corp. and later Intel Corp. Other threads are just as critical but less well known, like the electron and microwave tube industry of the 1940s and the tremendous growth in military aerospace efforts in the ’50s.
But one intriguing thread has been entirely forgotten, if it ever was really known. It connects the Nobel Prize–winning physicist William B. Shockley, the chemist and industrialist Arnold O. Beckman, the automation craze of the 1950s, and Shockley’s vaulting imagination of a robotic workforce. Shockley’s pursuit of that vision, and his desire to profit from his own ideas, led him to leave a successful career at Bell Telephone Laboratories to pursue a company of his own on the West Coast. There, Shockley forged a partnership with Beckman and created his silicon electronics lab—the region’s first—in Mountain View, Calif., thereby seeding the dramatic transformations that would eventually lead to the emergence of Silicon Valley as the global epicenter for silicon electronics.
The plot begins in 1948, a time of creative ferment for Shockley. That year, the 38-year-old Bell Labs researcher conceived the junction transistor, which quickly became the predominant type in the 1950s and 1960s. The first pocket transistor radios had junction transistors, as did the intercontinental ballistic missiles and bombers of the early Cold War. The invention cemented Shockley’s reputation as one of the foremost minds in the new world of semiconductor electronics.
Against that backdrop, then, what else was occupying Shockley’s thoughts? As it happens, 1948 was also the year that Shockley filed a patent application for a “Radiant Energy Control System”—basically a feedback control system that used a visual sensor. The design had nothing to do with transistors and everything to do with Shockley’s military work during the Second World War, especially his efforts to improve strategic bombing. He filed the application on 19 March, and it was quickly met with a patent secrecy order, remaining under wraps for a decade.
Figure 4 of the application makes clear the government’s concern: It depicts what appears to be a self-guided bomb falling on a factory. Housed in the nose of the bomb, Shockley’s guidance system would compare a series of photographic images—on a roll of film, for instance—with real-time images captured by a camera. Light entering the camera would pass through a frame of the film before falling on light-sensitive vacuum tubes. The better the match between the incoming image and the photographic frame, the larger the electrical signals. These signals, in turn, controlled servomotors that moved the fins of the bomb in a feedback loop. In his illustrated example, the film consisted of a series of aerial photographs of a military target taken at various altitudes. His control system would fly the bomb to that target.
Shockley also noted peaceful uses for his invention: facial recognition for access to buildings, identification of money in vending machines, and, intriguingly, “in factory production to sort items and inspect them.” This is perhaps Shockley’s first serious imagining of a machine, self-guided by feedback control and a visual sensor, to replace factory workers.
The idea of replacing humans with machines was nothing new, of course. From the Great Depression, through the war, and into the Cold War, the nexus of labor and manufacturing technology was a continual source of both innovation and conflict. When Shockley filed his patent application in 1948, there was widespread concern among the U.S. business elite that organized labor had become too powerful: Union membership had soared during the war, as had the number of strikes. Many in government and industry also worried that the United States lacked the industrial might of the Soviets. Clever, electronics-infused, self-guided machinery promised a solution to both concerns.
In fact, a word had recently been coined for precisely this solution: automation. Delmar Harder [PDF], a manufacturing leader at Ford Motor Co., used the word to describe a new class of machinery that could autonomously move materials from process to process along a production line. John Diebold, a graduate of the Harvard Business School, soon popularized the term and the idea with his 1952 best seller, Automation: The Advent of the Automatic Factory. In it, he described how new technologies coming out of the war—electronics, feedback controls, instrumentation, and electronic computers—could bring about an age of smart, adaptive, programmable factories that operated largely on their own. He believed this automation revolution was close at hand.
And yet, Diebold distanced automation from the idea of the humanlike robot. He asserted that such robots made little economic sense, even if they could be fashioned. William Shockley disagreed.
Two years after filing his now-secret patent application, Shockley met with Georges Doriot, a respected professor at the Harvard Business School and Diebold’s mentor. Doriot was also a pioneer in venture capital whose American Research and Development Corp. (ARDC) invested in high-tech spin-offs from MIT; its biggest winner would be Digital Equipment Corp.
Doriot and Shockley’s discussion turned to automation, a subject then gripping the MIT community following the publication of Norbert Wiener’s influential 1948 book, Cybernetics. Shockley saw that his control system could be a key technology in such a milieu. Doriot encouraged Shockley to think hard about the full implications of his invention for manufacturing.
Shockley wasted no time in doing so. By the start of 1951, he’d modified his designs and created what he called an “optoelectronic eye.” With it, he believed that a new class of machines could flexibly produce different products, not just handle continuous runs of fixed, standardized units.
Before formally recording his ideas, Shockley requested an extraordinary “modification agreement” to his intellectual property contract with Bell Labs. Shockley, like everyone else at the labs, had signed an employment agreement that gave the company the rights to all of his patentable inventions. Now he wanted Bell Labs to let him patent for himself his new optoelectronic eye for automated production.
After months of negotiation, Bell Labs and Shockley finally signed the agreement on 5 December 1951. But the new agreement gave the company an important out: It would last for just one year and was strictly limited to automation. During that time, Shockley could patent what he could—and claim any resulting benefits across the lives of those patents—but after it expired, it would be as if the agreement never existed.
With support from Doriot’s ARDC, Shockley completed his patent application for an “Electrooptical Control System.” Even before he was finished, however, he was feeling confident enough in his idea to write a remarkable memo to Bell Labs’ president, the physicist Mervin Kelly. The memo argued that Bell Labs should undertake a large, high-priority effort, led by Shockley, to create an “automatic trainable robot.”
“The importance of the project described below is probably greater than any previously considered by the Bell System,” Shockley declared. “The foundation of the largest industry ever to exist may well be built upon its development. It is possible that the progress achieved by industry in the next two or three decades will be directly dependent upon the vigor with which projects of this class are developed.” Even for the notably self-confident Shockley, this was forceful language.
“What is aimed at in this project is the substitution of machines for men in production,” he continued. He made it clear that he wasn’t talking simply about advanced machines for an “automatic factory.” He was talking about the robots of science fiction. Shockley’s trainable robots would perform the “manual operations of human operators” and “be readily modified to perform any one of a wide variety of operations.” Such robots could replace not just “semiskilled” industrial workers but also domestic workers in the home.
“A trainable robot will comprise ‘hands,’ ‘sensory organs,’ a ‘memory,’ and a ‘brain,’ ” Shockley wrote. First would be the “robot eye” that Shockley was about to patent. He added that he had “several ideas” for a “courageous direct attack on sense of touch devices...that can recognize parts and their orientation by feel.” That is, Shockley would give his robots first sight and then touch. Presumably, the memory, brain, and other senses would follow. Like Diebold, Shockley believed the automation revolution was imminent. With his robots, he predicted, “the increase in productivity of the nation might be...doubled in a decade or less.”
Shockley filed his robot eye patent application in November 1952, just before his modification agreement with Bell Labs expired. Ten days later, Shockley got his answer from Kelly: Bell Labs would not support the creation of a trainable robot.
Shockley responded with his feet. He soon took a leave from Bell Labs and would never fully return. He spent a year as a visiting professor at Caltech, making connections to the booming electronics, instrumentation, and military aerospace industry in the Los Angeles area. He had serious discussions with a number of firms to establish a transistor company of his own. None of the prospects satisfied. Shockley returned to the Pentagon, where as the director of research for the Weapons Systems Evaluation Group, he weighed in on issues such as how the U.S. military could best fight a nuclear war.
In December 1954, the U.S. Patent Office granted Shockley’s patent on the robot eye. Given Bell Labs’ rejection of his pet project, it might have seemed a hollow victory, and yet as the next few months showed, Shockley still believed that the robot eye might prove useful for industry.
Right after New Year’s in 1955, he received an invitation from Arnold O. Beckman to return to Los Angeles for a gala honoring Shockley and vacuum-tube pioneer Lee De Forest. At the event on 2 February 1955, Shockley was delighted to discover that Beckman—a Ph.D. chemist and prominent high-tech industrialist—was a fellow devotee of automation.
Born in 1900 and 10 years Shockley’s senior, Beckman was the son of an Illinois blacksmith. A tinkerer and inventor even as a youth, he’d studied chemistry and chemical engineering at the University of Illinois and went on to get a Ph.D. in physical chemistry at Caltech in 1928. Beckman took a hiatus from his Caltech studies to work at the newly formed Bell Labs, where he learned electronics and assisted in very early work on statistical quality control.
After finishing his doctorate, Beckman joined Caltech’s chemistry faculty. In the mid-1930s, he developed a pH meter, which combined chemistry and electronics in a simple, integrated instrument. Produced by a small firm in which he had a stake, Beckman’s pH meter was a runaway success. By the end of that decade, Beckman decided to leave his professorship, take control of the instrument firm, and devote himself to exploring the new world of electronic chemical instrumentation.
Through the 1940s Beckman’s firm enjoyed a string of other instrument and electronics breakthroughs, including his own invention of a hugely successful helical potentiometer—the Helipot—and his company grew dramatically. Beckman pushed his firm, renamed Beckman Instruments in 1950, to expand into automatic process control and electronic computers, and he became a devotee of automation and the automatic factory. His first report to shareholders, in October 1952, began with the slogan “Machines liberate men.”
Beckman’s pursuit of automation proved lucrative. By the end of 1954, sales had more than doubled, as had the value of the stock; profit approached $1 million, and the number of employees climbed to 1750. Caltech elected Beckman to its board of trustees, the first of its alumni so honored.
At their meeting at the Los Angeles gala in 1955, Beckman and Shockley were much impressed with each other. Both had science Ph.D.s from top schools, had studied at Caltech, had worked at Bell Labs, and possessed high-level security clearances. Each had era-marking inventions to his credit: Shockley, with his role in the first transistor and his conception of the junction transistor, had provided the foundation of the transistor industry, while Beckman, with his pH meter and ultraviolet spectrophotometer, had inaugurated the modern analytical instrument industry.
At the gala, Shockley promised to send Beckman his newly issued patent on the robot eye, which Beckman agreed to evaluate for possible use in his company’s automation efforts. This promise started a conversation that led, seven months later, to the two men signing a contract to create the Shockley Semiconductor Laboratory of Beckman Instruments Inc.
Shortly after the gala, Shockley sent Beckman a copy of the patent, asking for “any reactions you have about its potential usefulness.” Beckman, though, was focused on several corporate acquisitions and his company’s expansion into Europe. He sat on Shockley’s patent for nearly two months before passing it along to Jack Bishop, his right-hand man, asking for “R&D comments.”
Bishop in turn handed the patent to a trusted engineer, who returned a rapid and negative evaluation: “This system appears to me to be primarily of academic interest and…should not be given serious consideration at this time.” The engineer thought that in an industrial application, there would be easier and cheaper approaches than Shockley’s and that despite “the fact that the ideas are good,” a competitor could easily circumvent Shockley’s patent. It made no business sense, in other words. The engineer’s report made its way to Beckman, who held onto the R&D evaluation for another month and a half.
Finally, in mid-May 1955, Beckman responded to Shockley. Addressing Shockley as “Bill” and signing as “Arnold,” he informed the physicist that his engineering group had examined the patent for use in current or upcoming projects but that “it appears that there are no projects at this time.” His closing softened the blow a little: “It was a pleasure to see you again and I hope that our paths may cross frequently.”
Even as he awaited Beckman’s reply, Shockley was busy. During those months, he considered and then rejected an offer from Howard Hughes to lead a new consolidated semiconductor electronics organization at Hughes Aircraft. Shockley similarly explored then turned down offers to join the physics departments of the University of California, Berkeley, and Yale. All of these offers were extremely prestigious and lucrative. Yet they were also fundamentally conventional. Shockley sought a more radical change for himself. By June 1955, at the age of 45, he decided to leave Bell Labs, resign from his high-level Pentagon job, and exit his marriage. He wanted a new life and a company of his own through which he would be the primary beneficiary of his ideas.
At the end of July, Shockley called Beckman. He told Beckman about his desire to start a new company and made a new pitch: to take the chemical techniques for making silicon transistors recently created at Bell Labs and bring these new “diffused” silicon transistors to market. What’s more, he would develop automated means for the mass production of the transistors. Unlike his reaction to Shockley’s robot eye, Beckman’s response this time was immediate and enthusiastic: He asked Shockley to meet with him without delay. Two weeks later, on a Saturday, the two sat down for serious discussions.
Beckman was most likely attracted to Shockley’s new pitch for its fit with his overarching automation strategy. The high temperatures, strong vibrations, and caustic environments of chemical plants and petroleum refineries posed severe challenges for the vacuum-tube electronics then used in sensors and instruments. Rugged and reliable silicon transistors would therefore be invaluable for adapting Beckman’s electronic instruments to the control and monitoring of industrial processes. The U.S. military was also very interested in silicon transistors for their ruggedness and reliability.
After the meeting, it took Beckman and Shockley just a few weeks to hammer out a deal. In early September 1955, they signed an agreement creating the Shockley Semiconductor Lab. The agreement stipulated that the lab would remain a part of Beckman Instruments for two years, during which Shockley would have unfettered freedom to pursue new inventions in electronics. But the lab’s first activity, the two men agreed, would be the automated manufacture of diffused silicon transistors. Shockley insisted that Beckman give the effort his personal attention and that Beckman Instruments provide the lab with business, administrative, marketing, and sales support. After two years, Beckman would decide whether or not to make the lab into a separate company in which he and Shockley would have large equity stakes.
Shockley spent the fall of 1955 setting up his new lab. Although Beckman’s company was headquartered in Southern California, Shockley convinced him to locate the lab near Palo Alto, Calif. Shockley had grown up in Palo Alto, and his mother still lived there. Stanford University was nearby, and the area had a vibrant scene in electron-tube production for military and commercial markets, as well as nascent aerospace activities. San Francisco just to the north was a financial, military, commercial, and cultural center. And the weather was excellent.
Shockley drafted a press release to announce his new laboratory. It reveals his desire to assemble a team of virtuosi for semiconductor innovation and, in doing so, to give birth to a semiconductor community on the San Francisco Peninsula. “The opportunity to contribute to the growth of a new electronics community will attract men of imagination and initiative along both social and technological lines,” he wrote. “I hope to create a vehicle in which such men can make their maximum contributions and find their greatest satisfactions and rewards.” Though it’s unclear if the press release was ever made public, Shockley’s aspirations resonate with the image that Silicon Valley later attained.
Shockley assembled a team of “men of imagination and initiative,” including Gordon Moore, Robert Noyce, Jean Hoerni, Eugene Kleiner, and Jay Last. Then, in late 1956, Shockley shared the Nobel Prize in physics for his role in establishing transistor electronics. Despite these successes, Shockley’s lab soon shattered.
Shockley’s competitive streak and capricious management style were unfettered. He decided to abandon the diffused silicon transistor in favor of an invention of his own, the four-layer diode. The young researchers, including Moore and Noyce, became deeply concerned and dissatisfied, and they appealed to Beckman for help. They told Beckman that they still wanted Shockley as an adviser but that they needed a real manager. In the end, after getting assurances from Shockley and his friends in Bell Labs’ top brass that Shockley’s direction was essentially sound, Beckman stayed the course: Shockley would continue to lead the lab.
In response, at the end of 1957, Moore, Noyce, and six others quit and created a new company, Fairchild Semiconductor, just a couple of kilometers away. In Silicon Valley’s first silicon electronics spin-off, the “traitorous eight” (as Shockley reputedly called them) focused on making the diffused silicon transistor that Shockley had abandoned. They succeeded within a year and over the next several years introduced important new transistors and integrated circuits based on the “planar process,” a breakthrough manufacturing technology that allowed circuits to be chemically printed in silicon substrates. As planar silicon transistors and microchips replaced vacuum tubes, “electronics” and “semiconductor electronics” became increasingly synonymous.
From the 1960s into the 1980s, further spin-offs from Fairchild and related start-ups in semiconductor equipment and materials populated the San Francisco Peninsula. Counting the spin-offs of these spin-offs, Fairchild’s descendants numbered in the hundreds. In 1968, Moore and Noyce created one of these “Fairchildren”: Intel Corp. First pioneering memory chips and then microprocessors, Intel eventually became the world’s largest semiconductor company.
Seen in this light, Shockley Semiconductor was incredibly productive. But for its founders, Beckman and Shockley, the lab was a failure. It never produced a silicon transistor and found few takers for the four-layer diode. Finally, in 1960, Beckman sold the operation, and it closed several years later. Shockley retreated to a professorship at Stanford, and Beckman would forever regret that he had not retained as his employees the dissidents who founded Fairchild and Intel.
Until his death in 1989, Shockley never returned to his grand vision of the trainable robot, even as others made industrial robotics real. In the late 1950s, George C. Devol started the robotics company Unimation, whose transistorized robotic arm, the Unimate, entered service in the early 1960s to move hot metal castings in a General Motors plant. Similar programmable robotic arms proliferated during the 1970s, as improvements in microchips made computing power more affordable. Then, in the 1980s, robust machine vision—successor to Shockley’s “eye”—was added to these robotic arms.
The progression of robotics continues, to the point where household robots, weaponized drones, intensely automated factories, and self-driving cars now seem almost commonplace. In the end, Shockley did prove instrumental in the realization of the robots he had imagined, but not in the way that he had hoped. The microchip industry that emerged, in part, from Shockley’s lab enabled the rise of robotics, but he himself played no role in the robotics revolution.
For more about the author, see the Back Story, “Connecting the Documents.”
This article originally appeared in print as “Shockley’s Robot Dream.”